ABSTRACT Stroke is one of the leading causes of death and disability in the United States and other developed countries. About 800,000 people in the United States have a stroke each year. The incidence of stroke is expected to increase (due to diabetes, an aging population and a rise in obesity), which is a major public health concern and great economic burden to society. Early identification of people at risk of stroke is important as it allows for better planning and implementation of preventative strategies. A number of risk prediction models have been proposed for stroke using common biomarkers such as lipids, and conventional risk factors including age, gender, history of hypertension and smoking. However, these models do not adequately capture a substantial proportion of persons at high risk of stroke they lack precision (best accuracy is up to 50% by Framingham risk score) in risk estimation and have other limitations. Therefore, it is a necessity to improve the accuracy of a stroke risk prediction model. Therefore, we aim to develop a novel model using retinal vessel features and conventional risk factors which will provide higher accuracy on stroke prediction than the existing models and provide better preventative strategies for the people at risk of stroke.